I realised that last year was the 15th anniversary of my first published program, a game called Monster Builder on the BYOND platform, which I made in early 2002. Really it was just a small modification of an existing game, whose name I can’t remember. In 2004 I gave a copy of the source code to someone, they posted it online, and I got mad and deleted the original game. I then put it back online in 2006.
You can tell it’s from a long time ago because I still capitalised “Barrucadu”. In fact, that account on BYOND is the source of the name: it was a joint account for me and my brother. He wanted to be a marine biologist at the time but, alas, couldn’t quite spell “barracuda”.
- I implemented the feedback I got on my operational semantics last week, and that section is now looking pretty good.
I was still on support Monday and Tuesday this week:
I had another data.gov.uk harvesting problem to fix. That’s quite the ordeal for a Monday morning.
We got a support ticket to move some content from Natural England to the Rural Payments Agency (RPA), and the ticket came with a “preliminary list” of pages needing changes. I thought “preliminary, that means incomplete—I wonder if they just want all Natural England content changing?” So I asked:
There are 1524 documents in Whitehall Publisher where Natural England is the lead organisation, and 197 where Natural England is a supporting organisation. This includes all documents, even those which have never been published (the only edition is a draft, for example), and which have been withdrawn or superseded. I have attached a list.
Would you like them all changed to the RPA, or just the ones in the your spreadsheet? I can do this for you today.
Michael Walker | GOV.UK Developer
I’m sure you can see where this is going. RPA guy said to go ahead and change them all. I made the change. The next morning we got an urgent support ticket from Natural England asking why all their content had been moved. Whoops.
After getting off support, I started to look into some memory issues we’ve been having with our asset-manager and publishing-api applications; usage had roughly doubled since we started using AWS X-Ray. In the asset-manager, I found that enabling X-Ray caused each thread to use an additional ~100MB resident memory (from 87 to 192): given that it’s just recording the target and duration of HTTP requests and sending them to a daemon on localhost, this seemed crazy!
I tracked down the problem to enabling tracing of requests made using the aws-sdk gem. Tracing of regular net_http requests is, ironically, fine. So we disabled tracing of aws-sdk requests and raised an issue on aws-xray-sdk.
I finished off the work from “Weeknotes: 001” on making transition pull its stats from an S3 bucket, it just needs deploying. Once we’re satisfied that the new code is doing the right thing, we can delete 5GB of logs stretching back 6 years, and also the three machines (one in each of production, staging, and integration) which exist solely to process them.
I asked why something was the way it was, and got linked to the Trello card I wrote about why it was the way it was. I then changed the way it was, so it no longer is as it was.
I discovered how easy it is to make really tasty roasted onions (thick-slice the onions; mix with olive oil, sea salt, and pepper in a roasting tray; perhaps add some sliced garlic and/or worcestershire sauce; put in oven on 200C for 45 minutes, stirring in the middle). I can never go back to fried onions.
One of the RAM sockets on my desktop computer’s motherboard died on Thursday. The symptoms: the computer locked up, and when I power-cycled it, both Windows and Linux panicked during boot. Windows gave a blue screen with a message about an unexpected interrupt, Linux varied but most consistently complained about a general protection fault.
This was a bit tricky to figure out, because the symptoms of a dead RAM socket are much like the symptoms of a dead RAM stick, until you think you’ve isolated the problem and it then breaks in a new way. So I’m down from 16GB of RAM to 12GB, which isn’t much of a problem, but more worryingly my motherboard is possibly beginning to fail.
I contributed a small documentation fix to pleroma.
I was flipping through my academic reading list, which has been largely untouched since I moved to London, and came across Can automated pull requests encourage software developers to upgrade out-of-date dependencies? At GOV.UK, the answer is definitely a resounding “yes”.
I only thought of this section on Saturday evening, and I have the memory of a goldfish, so it’s a bit sparse.
There’s some good advice here, particularly about reading error messages. A surprising number of people don’t.
I really like this distinction because I’m in the latter camp, and never really thought about the former. My attitude toward types is “of course you must know what type something is, otherwise how can you do anything with it?”, which naturally leads to a preference for static typing over dynamic typing.
This 404 page is a blast from the past.